SPC Workshop #1: Process Simulation, PROSIM
Introduction
In this workshop, a sampling experiment will be performed with the aid of PROSIM, an
MS EXCEL-based Visual Basic program that generates samples of data from a fictional process.
X and R control charts can then be constructed from the samples to assess the process behavior
history.
Learning Objective
The objective of this computer workshop is to provide you with practice in identifying statistical
signals on control charts, and the use of these signals to deduce the underlying process behavior.
Description of the Computer-Based Process Simulator, PROSIM
PROSIM randomly selects a process history from one stored in a process history library. A
process history is a chronological description of the distributions/behaviors used to generate the
sample data. PROSIM draws 125 subgroups of size n=5 using the information associated with
the selected process history. A grouping of samples within a history may have one of the
following normal distributions associated with it: N(0,22), N(2,22), N(-2,22), N(0,12), N(0,42),
N(1,12), and N(1,12). Alternatively, a group of samples may be linked to an inappropriate
sampling method (stratification or mixing) or the process fault of overcontrol. For example, a
process history could be
Samples 1-28: N(0,22) Samples 78-84: N(0,42)
Samples 29-37: N(-2,22) Samples 85-109: Overcontrol
Samples 38-45: N(2,22) Samples 110-118: N(0,42)
Samples 46-69: Stratification Samples 119-125: N(0,22)
Samples 70-77: N(0,12)
There are three rules that PROSIM uses when generating the process data:
1. At least the first 25 samples of size n=5 are taken from the distribution N(0,22).
2. Shifts in the mean and/or variation are sustained for at least five consecutive subgroups.
3. Stratification, mixing, and overcontrol are sustained for at least 20 consecutive subgroups.
Overcontrol is simulated through changes to the mean of a normal distribution with a standard
deviation of 2 using the following rules:
1. If the sample mean, X, of the most recently collected sample is within a narrow range about
the target value of zero, no action is taken.
2. If the X is above the target value (and outside the narrow range), the true mean for the next
sample is reduced.
3. If the X is below the target value (and outside the narrow range), the true mean for the next
sample is increased.
Once PROSIM has generated the sample data inside the EXCEL spreadsheet, control charts may
be constructed for the sample data using EXCEL, or, the data may be exported to another file for
subsequent control chart construction via third-party software. The charts may then be
interpreted for statistical signals.
Assignment
Ultimately, you should be able to draw conclusions about the specific distributions/behaviors
that were used and the time-period over which they were present.
At least four tasks must be completedt:
1. Run PROSIM to simulate the collection of 125 samples of size n=5 from the fictional
process.
2. Construct and interpret control charts for the data collected from the fictional process.
3. Study the charts and determine where the statistical signals are and what they are saying
concerning shifts in the mean and/or the standard deviation of the process.
4. Prepare a report that summarizes what you have learned about the process.
The report, describing what you have learned about the process, must include at least the
following:
1. An executive summary, which in approximately one page summarizes the contents and
conclusions of your report.
2. A body of the report, which includes the following components:
a. An introduction, which briefly describes the concepts and methods relevant to the
interpretation of statistical signals on control charts and indicates the methodology
employed in identifying the history of the fictional process.
b. A procedures section, which describes the methodology used to study the process.
Remember that the identification procedure you are undertaking involves two basic steps:
(i.) identifying statistical signals indicating the presence of a special cause, and
(ii.) identifying the specific nature of that cause (i.e., how the process is being affected).
The logic used in deciding what to do at each step should be spelled out carefully. You
should clearly state how you arrived at any conclusions. In this section it will clearly be
necessary to refer to the control charts themselves, which should be contained in the
appendix to the report.
3. A conclusions section, which summarizes the special-cause nature and time period over
which you believe each special cause was present. Remember, you can conclude that a
special cause is present only when the control charts so indicate. Your conclusions may be
nicely summarized with a figure showing how the process changes over time.
4. An appendix section, which contains the PROSIM raw data and the control charts that were
generated during the course of your investigation. Note that these control charts should
clearly indicate (i.e., with circled points) all the statistical signals on the charts.